Kartika Candra Kirana, Frans Achmad Hendra Winata, Wahyu Nur Hidayat, Muis Muhtadi, Heru Wahyu Herwanto, Slamet Wibawanto
Previous research has used YOLO as a computer vision technique to recognize sign language translation systems. Nevertheless, there is no research focusing on optimizing the model's hyperparameters. However, hyperparameters are crucial keys to optimizing the Yolo model. The goal of this study is to optimize YOLOv8 model hyperparameters to develop Indonesian sign language learning media. The best possible hyperparameter configuration was found through testing, including YOLOv8m pre-trained weights, 150 epochs, the Stochastic Gradient Descent (SGD) optimizer function, and a batch size of 16. The result shows 100% for Precision, Recall, and F1 Score for 4 real-time videos, as well as a measure indicating that the proposed model could recognize sign language movements as learning media. © 2024 IEEE.
Universitas Negeri Malang, Department of Electrical Engineering, Malang, Indonesia
